In a plant system such as a chemical plant or an oil refinery plant, dust and the like may adhere to the inner wall of the piping due to aged deterioration, resulting in problems in the operation of the plant. For this reason, inspection and cleaning of piping is performed with a regular interval. The interval for the inspection and cleaning is generally determined according to empirical rules, and it is not clear whether the interval is appropriate or not. Therefore, in order to optimize the interval for the inspection and cleaning and to reduce maintenance cost, detection of occurrence of the aged deterioration is needed.
For example, an invariant relation analysis described in PTL 1 is known as a technique for detecting an abnormality of a plant system. In the invariant relation analysis, a correlation model indicating statistical relation between metrics (performance indicators) of the system is generated, and a difference between the behavior of the system and the correlation model is detected as a system abnormality.
As a related technique, PTL 2 discloses a technique for estimating a cause of failure by using information about correlation in which correlation destruction is detected during occurrence of failure, in the invariant relation analysis. PTL 3 discloses a technique for detecting a sign of failure of a system by using distributions of correlation destruction in normal state and in abnormal state, in the invariant relation analysis.